README.md

metadata

language:
  - fr
  - en
license: apache-2.0
tags:
  - medical
  - continual-pretraining
  - instruction-tuning
  - cpt+sft
  - causal-lm
  - question-answering
base_model: BioMistral-7B
model_type: causal-lm

BioMistral-7B-CPT-SFT (CPT+SFT)

Model description

This checkpoint applies two-stage domain adaptation:

1. CPT on unlabeled medical text to adapt representations on Dr-BERT/NACHOS corpus , then
2. SFT on labeled medical QA to optimize task performance (MedInjection-FR/ALL) .